The original title is "A Practical Workflow for Contributing to a Large, Structured Codebase"

Short summary

A developer shares a structured workflow for using AI agents like Claude Code to contribute to large codebases: read specs and conventions first, respect architectural dependency order, follow repository rules for DTOs and citations, verify with tests and runtime checks, and document any deviations. The process emphasizes small focused changes, honest failure reporting, and clean finishes with status updates and ADRs. It aims to reduce regressions and ensure AI-generated code integrates naturally.

• •Five-step workflow: understand requirements, respect dependency order, write code per repo rules, verify with tests, finish cleanly
• •Key rules: generated DTOs only, source citations for ported behavior, minimal meaningful comments
• •Emphasizes honest failure reporting and documenting convention deviations with ADRs

Generated with AI, which can make mistakes.